1

Folding Dynamics of Single GCN-4 Peptides by Fluorescence ResonantEnergy Transfer Confocal Microscopy

Yiwei Jia1, David S. Talaga1, Wai Leung Lau2, Helen S. M. Lu2, William F. DeGrado2,Robin M. Hochstrasser1

1Department of Chemistry and 2Department of Biophysics and BiochemistryUniversity of Pennsylvania, Philadelphia, PA 19104

AbstractWe have prepared a bichromophoric crosslinked variant of GCN4-P1 for single

molecule fluorescence energy transfer experiments (GCN4-Pf). The folding and

unfolding fluctuations of single GCN4-Pf molecules are measured in a two channel

confocal microscope with which donor and acceptor fluorescence trajectories are

measured simultaneously. The energy transfer efficiency is thereby determined and its

probability distributions as a function of added denaturant [urea] are calculated. The

distributions indicate that single molecule GCN4-Pf is in dynamic folding equilibrium

with the position of the equilibrium being altered by the concentration of urea.

2

Introduction

Single molecule measurements are now moving from the realm of technological

demonstrations to making contributions to new understanding of chemical systems

particularly those that are microscopically inhomogeneous and exhibit dynamics over

many time scales.[1-9] Using single molecule detection, it is now possible to follow the

evolution in time of individually selected members of an equilibrium ensemble. This

trajectory can be used to evaluate rates, rate constants, and distributions of other

properties. On the other hand, bulk measurements usually give only ensemble averaged

value of the molecular property in question.

Proteins and other biological assemblies exhibit microscopic structural

heterogeneity and are therefore of particular interest for single molecule studies. The

structural fluctuations of proteins can result in folding and unfolding of the primary

sequence of amino acids between a well defined three-dimensional native structure and a

broadly distributed set of denatured structures. [4,10-14] In one example, fluorescence

spectral fluctuations were attributed to dynamics amongst protein substrates in single

molecule experiments.[6] In another study, protein conformational dynamics was cited as

the origin of fluorescence intensity and polarization fluctuations of single Staphylococcal

nuclease.[4]

This paper concerns the study of a peptide derived from the yeast transcription

factor, GCN4. The DNA binding domain of this protein includes a sequence that forms a

short segment of a two-stranded coiled coil [15,16], as shown in figure 1. Since their

discovery, coiled coils have provided very simple model systems for the study of the

folding of water-soluble proteins.[12,13,17] A peptide spanning the coiled coil of GCN4

3

(GCN4-P1) has been shown to form a cooperatively folded helical dimer. [12,13,18-22].

This peptide is an excellent system for studying protein folding because it is quite simple,

and yet contains a well-packed helix/helix interface, as found in globular proteins. It has

been shown to exist in a two-state equilibrium between unstructured monomers and fully

alpha-helical dimers12. The alpha-helical secondary structure and the double-helical

folded structure apparently form concomitantly.[13,17-20,23] To simplify the folding

reaction, Kim and coworkers have introduced a covalent disulfide tether between the

two-peptide chains. The crosslink stabilizes the proteins towards thermal denaturation,

although the peptide continues to fold in an apparent, two-state equilibrium. [24]

One purpose of this study is to investigate the microscopic features of a

macroscopically observed kinetic model. GCN4-P1 exhibits two-state folding kinetics

when in bulk solution [17,20], although a folding intermediate has occasionally also been

observed for some coiled coils [13,19,23]. It has been shown, however, that macroscopic

averaging can create effective two-state kinetics in the bulk even when there are multiple

unfolded states of the protein.[10]

Single molecule experiments are sensitive to mechanistic heterogeneity. Since

single molecule experiments are immune to ensemble averaging, it is possible, in

principle, to distinguish between a pathway-based folding mechanism and a landscape

multi-path mechanism. If a protein has a single folding pathway, then the distribution of

observed kinetic rates should be narrow. Conversely if there are multiple pathways there

would, in general, be multiple rates associated with folding and an appropriate

distribution of rates would be observed in a single molecule experiment.

4

Modifications to GCN4

The GCN4-P1 variant employed in this study, designated GCN4-Pf, has the

sequence

GGRMKQLED10KVEELLSKDY20HLENEVARL30KKLVGERGG40CGEEEEE. Its

design incorporates several distinct features (Fig. 1): Asn-18, which forms a buried

hydrogen bond with Asn-18′ of a neighboring helix in GCN4-P1, has been changed to

Asp in GCN4-Pf. The presence of aspartic acid at the helix/helix interface of the peptide

introduces a "pH switch," which destabilized the peptide at high pH, allowing the

modulation of stability with pH as well as urea.[25] Five glutamic acid residues were also

appended to the C-terminus, providing a flexible appendage to allow electrostatic

adsorption of the peptides onto a positively charged surface for single molecule studies.

A Gly-Gly-Cys-Gly tetrapeptide was inserted between GCN4-P1 and the polyglutamic

acid tail. Crosslinking of the two helices by a disulfide bond formation at the Cys residue

provides an unimolecular-folding situation. This simplification is of particular importance

given the low concentrations (< 10-11M) used for single molecule experiments where the

likelihood of observing an “on” bimolecular folding event would otherwise be

vanishingly small. The Gly residues serve as a flexible linker between GCN4-P1 and the

polyglutamic acid tail.

In order to probe conformational fluctuations in the folded and unfolded states we

synthesized GCN4-Pf with a fluorescence donor and acceptor pair attached to the N-

termini. Initially, two different reduced cysteinyl peptides were prepared with the same

peptide sequence but different N-terminal fluorophores that form an energy transfer pair.

Texas Red-X (TxR) behaves as the energy acceptor while 5-carboxyrhodamine 6G (R6G)

5

behaves as the energy donor. These fluorescently labeled monomers were first reduced,

then recombined under oxidative conditions to form the desired heterodimeric peptide.

Experimental

Materials

Abbreviations used: APS, aminopropylsilanyl; EDTA, ethylenediaminetetraacetic

acid; HOBT, 1-hydroxybenzotriazole; Fmoc, 9-fluorenylmethoxycarbonyl; MES, 2-(N-

morpholino)ethanesulfonic acid; KCl, potassium chloride; R6G, rhodamine-6G; TxR,

Texas red; TFA, trifluoroacetic acid; Tris, tris(hydroxymethyl)aminomethane; MRE,

mean residue ellipticity.

(3-Aminopropyl)trimethoxysilane was purchased from Lancaster synthesis, Inc..

Reagents for peptide synthesis were purchased from Advanced Chem Tech, Inc. HOBT,

piperidine, and picric acid were obtained from Aldrich. Solvents were obtained from J. T.

Baker. Tris, EDTA, KCl and TFA were purchased from Sigma. Amine reactive dyes

were purchased from Molecular Probes: R6G #C-6127; TxR #T-6134)

Synthesis and Characterization of GCN-Pf

The GCN4-Pf polypeptide was synthesized using a standard solid-phase Fmoc

method on a PE Applied Biosystems 433A peptide synthesizer. The fluorophores were

coupled to the N-terminal amino group following assembly of the peptide on the solid

support. The N-terminal Fmoc group was deprotected by 20% piperidine to generate a

free amino terminus, and then coupled with approximately 1 equivalent of Texas Red-X

N-hydroxysuccinimide ester (or 5-carboxyrhodamine 6G N-hydroxysuccinimide ester)

and HOBT, 10 equivalents of diisopropylethylamine in dimethylformamide. The reaction

was conducted under argon in the dark for 48 hours. Following cleavage from the resin

6

with trifluoroacetic acid with thioanisole, ethanedithiol and anisole in the 90:5:3:2 (v/v),

the peptides were purified to homogeneity by reverse-phase HPLC using a Vydac C18

column (22x250 mm 15-20 micron particle size) running a linear acetonitrile-water

gradient (with segments of 0.25% by volume increase per minute of acetonitrile) in the

presence of 0.1 % TFA. If necessary the Cys residue was first reduced to the thiol by

reaction with tris[2-carboxyethyl]phosphine.

The GCN4-Pf heterodimer was prepared by cross-linking the individual 5-

carboxyrhodamine 6G and Texas Red labeled peptides via a disulfide bond. Equal

amount of the two peptides (l.1 mg/ml) were reacted in 0.25 mM (reduced/oxidized 1:1

ratios) glutathione, 0.1 M KCl, 0.1 M Tris pH 8.5 and 1 mM EDTA. Cysteine oxidation

resulted in the formation of two different homodimers and the desired heterodimer.

Peptide mixtures were fractionated by reverse-phase HPLC using a Vydac C18 column

(22x250 mm 15-20 micron particle size). The peptides were eluted by means of a linear

gradient from 33 to 45 % of acetonitrile in TFA 0.1 % over 52 minutes; elution was

monitored at 222 nm. The identity of the heterodimer was confirmed by MALDI-TOF

mass spectrometry using a Perspective Biosystems Voyager DE-RP BioSpectrometry

Workstation. A singly labeled GCN-4 with donor only was prepared by a modification of

this procedure.

CD spectroscopy of GCN4-Pf heterodimer in solution

Circular dichroism (CD) spectra of fluorescently labeled GCN4-Pf was measured

in 10 mM MES, pH 6.1 at a total peptide concentration of approximately 2 µM in a 1 mm

pathlength cell. The peptides showed the double minima at 208 and 222 nm

characteristics of alpha helices.

7

Aminopropylsilanization (APS) of quartz and glass surface

Clean glass or quartz slides (2 x 1 x 0.4 cm) were amino-silanized by reaction

with 0.1% (v/v) (3-aminopropyl)trimethoxysilane in hexane, providing a surface that was

positively charged at neutral pH or below. The reaction was allowed to proceed for one

hour at room temperature. To determine surface density of free amine on the modified

quartz, the plates were immersed in 0.1 M picric acid in dichloromethane, and then

washed with the same solvent. This procedure resulted in formation of chromophoric

picrate salts of the aminopropyl groups on the derivatized surface[26]. The slides were

then placed inside a cuvette filled with dichloromethane. The quartz slides were

maintained in vertical position by two custom-made Teflon spacers. The absorbance at

358 nm was used to estimate the surface density of amines, using an extinction

coefficient of picrate of 14,500 M-1cm-1. The free amino form of the derivatized surface

was regenerated by washing the slides with 5% DIEA (v/v) in DCM and then DCM. The

amount of picrate eluted into solution in this manner agrees well with the amount

detected directly on the surface. The quartz slides were used within 24 hours. The

degree of modification thus determined was approximately one amine per 20 Å2,

indicative of a single monolayer of well-packed monomers.

For single molecule studies, microscope cover slips (Fisher, No. 1.5) are modified

as above. GCN4-Pf is adsorbed to the surface by applying 30 µl of a 1.0 nM solution of

GCN4-Pf in 10 mM MES pH 6.1 buffer for 5 minutes. The surface is then washed three

times, then loaded with the same buffer containing the desired urea concentrations.

8

Circular dichroism of monolayers of GCN4-Pf on APS-modified quartz

GCN4-Pf heterodimer was dissolved in 10 mM MES pH 6.1 and then transferred

to a 1 cm pathlength quartz cuvette containing four APS-modified quartz slides held in a

vertical position by two custom-made Teflon spacers. The peptide was added at a

concentration of 4.6 µM, providing an approximately 5-fold excess of the peptide relative

to the amount required to form a uniform monolayer of the peptides in a vertical

orientation. The incubation time was one hour. The quartz slides inside the cuvette were

rinsed once prior to filling the cuvette with known volume of 10 mM MES pH 6.1 buffer.

Circular dichroism spectra were acquired on an Aviv 62A DS or 62A circular dichroism

spectrophotometer. Far-UV spectra were scanned from 300 to 195 nm. The amount of

surface-adsorbed peptide was calculated from its absorbance at 520 nm, using an

extinction coefficient of 108,000 mmole/cm2 (as given by Molecular Probes, Inc). The

observed density of peptide on the surface was 8 nm2/dimeric molecule. By comparison,

the cross-sectional area of the GCN4 coiled coil in crystals is 6.6 nm2/dimeric

molecule.[15]

As compared with the sample in homogeneous solution, the immobilized GCN4

peptide shows a similar CD spectrum, except the minimum at 208 nm is of lower

intensity (Figure 2). This finding suggests that the helices may be oriented perpendicular

to the quartz and parallel to the incident UV light, based on the known polarization of this

π-π* transition.[27]

A urea denaturation curve for surface-adsorbed GCN4-Pf was determined by

monitoring the change in [θ]222 as a function of urea concentration. At each urea

concentration, data were collected and averaged for two minutes. To facilitate mixing of

9

the aqueous contents, only two peptide-covered quartz slides were used in the urea

denaturation experiment. The data in Fig. 3 represent the average of three experiments.

The urea-induced denaturation of the peptide exhibits an unfolding transition midpoint

around 3 M urea. This value is similar to that observed in bulk solution (data not shown),

indicating that the quartz surface did not greatly perturb the peptide’s conformational

properties.

Single Molecule Apparatus

The home-built inverted scanning confocal microscope has been described

before.[9] To excite the single molecules we used a 76MHz mode-locked Nd:YAG laser

(Coherent Antares) frequency doubled to 532 nm. An average excitation power of 0.5

µW is coupled into the microscope by a dichroic mirror (550 DCLP, Chroma). The

excitation light is made circularly polarized by a λ/4 waveplate. The reflected and

scattered 532 nm light is removed by a 532 nm holographic filter (Notch Plus, Kaiser).

The fluorescence photons of R6G and TxR are separated by a second dichroic mirror

(580 DCLP, Chroma). Further rejection of background and crosstalk is achieved by

filtering the fluorescence of R6G with a band pass filter (540-580 nm, Chroma) and a

long pass filter (OG550, Schott). Similarly, a long pass filter (OG590, Schott) and a band

pass filter (600-660 nm, Chroma) are also used for the detection TxR Fluorescence. Two

Avalanche photodiodes (EG&G) provide simultaneous high quantum yield detection of

single photons from each of the fluorescent dyes. Fluorescence images of GCN-4 labeled

with both R6G and TxR were recorded simultaneously by scanning the sample stage

(area: 36 µm2). Photo-bleaching curves of R6G and TxR were simultaneously recorded

with an integration time of 0.98 ms.

10

Results

Macroscopic Characterization of GCN4-Pf in Solution on APS-quartz

The goal of the present paper is to measure conformational fluctuations of GCN4-

Pf in the folded and unfolded states, as well as under conditions where it is in dynamic

equilibrium between these two conformational ensembles. We therefore first measured

the folding equilibrium of GCN4-Pf, both in solution and adsorbed to APS-quartz. The

CD spectrum of GCN4-Pf (Figure 1) in aqueous solution and on APS-quartz is typical of

that of α-helical proteins. Preliminary solution experiments showed that the folded

conformation of GCN4-Pf was increasingly destabilized as the pH was increased from 3

to 8. The current studies were conducted at pH 6.1, because this value provided a

marginal degree of stability favoring the folded form.

Figure 3 illustrates a urea denaturation curve for surface-adsorbed GCN4-Pf, as

monitored by CD spectroscopy. The curve is similar to that observed for the peptide in

free solution, and shows a midpoint near 3.0 M urea. The urea-induced denaturation

curves of small proteins are generally analyzed assuming a two-state equilibrium

involving a folded and unfolded ensemble of conformers. The free energy for this

process, ∆Gfold, has been shown to scale approximately linearly with respect to the

concentration of urea. Application of this model to the data in Fig. 3 indicated that the

value of ∆Gfold (extrapolated to 0 M urea) was approximately –1.6 ± 0.4 Kcal/mol

(midpoint = 2.8 ± 0.3 m urea), although those values are somewhat uncertain because the

folded and unfolded baselines were not well resolved. Thus, GCN4-Pf is approximately

90% folded in the absence of urea at pH 6.1.

11

Images

Single molecule fluorescence images of GCN4-Pf at pH 6.1 in the absence of urea

are shown in Fig. 4. With 532 nm excitation, mainly the R6G is excited but the TxR

channel shows significantly more fluorescence spots. In addition, the intensity of the

emission of TxR is stronger than R6G. These results are consistent with the protein

system being mainly in a folded state at pH 6.1 in the absence of urea; the donor-to-

acceptor distance is relatively close (Fig. 1), and the energy transfer efficiency is high

under these conditions. Fig. 5 shows single molecule fluorescence images at pH 6.1 and

7.4M urea. In this case, the R6G channel shows significantly more emission, indicating

less effective energy transfer and an unfolded condition.

Dual Channel Detection

The overlap between the donor and acceptor emission spectra influences the

signals seen in each channel. To correct for this cross-talk we note that the overall signal

measured at each detector (SD and SA) is a linear combination of true donor and acceptor

intensities (ID and IA) and the observed background for each channel (BD and BA).

SD = C DD ID + C DA I A + BD, SA = C AA IA + C AD ID + BA

IA=

C DD SA− BA( )− C AD SD

− BD( )C DD C AA

− C AD C DA

, ID=

C AA SD− BD( )− C DA SA

− BA( )C AA C DD

− C DA C AD

The coefficients Cij represent the amount of i (i=A,D) signal reaching channel j

(j=A,D) and are determined experimentally using two different methods. We used the

fluorescence of singly labeled GCN4-Pf molecules to measure the cross-talk between

channels. During the course of the measurements of the GCN4-Pf we also measured the

cross-talk after the acceptor molecule became bleached. These measurements allow the

individual trajectories to be corrected to obtain signals proportional to the donor and the

12

acceptor fluorescence intensities. There is no appreciable leakage of the acceptor signal

into the donor detection channel.

Types of trajectories

Fig. 1 illustrates typical single-molecule recordings of the variation in

fluorescence intensity with respect to time for GCN4-Pf. We observe a number of

different types of trajectories. The trajectories in figure 6 are representative of those

acquired for GCN4-Pf. The trajectory in figure 6(a) is typical [i.e., occurs ca. 70% of the

time] of those providing strong evidence for energy transfer between the donor and

acceptor. The acceptor signal dominates until it photobleaches to generate a state that can

no longer act as an excitation energy acceptor. At that time, the donor signal jumps to its

non-perturbed level indicating no energy transfer. We interpret the abrupt termination of

the signal as photobleaching; it should not be confused with a folding/unfolding event.

The abruptness and completeness of this transition indicates that it is a single molecule

event. Another class of trajectories [ca. 25%] is shown in Figure 6(b), in which, within

the time resolution of the experiment, the acceptor and donor bleach simultaneously.

Sometimes the donor bleaches first [ca. 5%] and we see a reduction of the acceptor signal

to the level of its direct excitation, as in Figure 6(c). Figure 7 shows a single molecule

trajectory on an expanded time scale which more clearly shows the anticorrelated

fluctuations in the donor and acceptor signals. The anticorrelated signal was absent from

GCN4-Pf molecules labeled with only the donor, indicating the detection channels are

uncoupled. Furthermore, the mean squared signal fluctuations (excluding shot noise)

were ca. 6 times less for the single labeled peptide. Thus, most of the variance in the

13

trajectories arises from variations in the donor/acceptor distances, the angles between

donor and acceptor transition dipoles and the orientation of the dyes relative to the

surface.

Sources of fluorescence intensity fluctuations

Single molecule fluorescence intensity fluctuations in our experiments can arise

from a variety of sources. We are focused on extracting the fluctuations related to

stochastic motions in the folded and unfolded states, as well as the dynamics of

interconversion between those two states. Photophysical processes that may also give rise

to intensity fluctuations include time-dependent shifts in the fluorescence

spectrum[28,29], transitions in and out of non-fluorescent states of the system[30],

including triplet states[31,32] and irreversible photobleaching[1]. We eliminated the

contributions of some photophysical processes. The GCN4 labeled with only R6G or TxR

did not exhibit any cross correlations of the signal between the two channels, indicating

that fluctuations of the fluorescence spectral positions are not contributing to this signal.

Triplet lifetimes in the presence of O2 should be on the order of microseconds. We

deliberately avoided the use of an O2 scavenging system[2] to minimize complications

from transitions to triplet states that might otherwise complicate our interpretation of the

millisecond signal fluctuations. Finally, following photobleaching, the trajectories exhibit

long-lived non-fluorescent states; these latter portions of the trajectories were not

included in our analysis.

Angular motions of the transition dipoles of the probes R6G and TxR can also

contribute to the fluctuations. Because we use circularly polarized excitation and the

detection system is not polarization sensitive, the variations in the azimuthal angles, Φ,

14

do not influence the signal intensity. However, variations in the colatitude, θ, of both

donor and acceptor can affect the measured intensities. In independent experiments on

R6G or TxR bound to a GCN4-P1 peptide we measured the decay of the fluorescence

anisotropy. In both cases, the anisotropy of the probes relaxed on a subnanosecond time

scale exhibiting an order parameter of ca. 0.5. The total anisotropy decayed in 2-3 ns.

These results indicate fast rotational motion of the probes relative to the peptide and

significant rotational averaging of the energy transfer signals is occurring on the

nanosecond time scale. The relative signal detected from a radiating dipole having

colatitude θ, is within a few percent of sin2θ(1 + sin2θ) for our microscope, when the

excitation is circularly polarized[33]. Therefore the relative signal intensities from donor

and acceptor depend not only on the angles involved in the dipole-dipole interaction that

form part of the energy transfer efficiency, but also on the colatitudes θA and θD.

Therefore the signal will also contain contributions from fluctuations in these angles.

Calculation of the energy transfer quantum yield

The corrected count rates are used with the donor and acceptor extinction

coefficients to determine the quantum yield for energy transfer.

ID = εD

ØD 1 − ØET( )1 − ØET 1 − ØD( )

IA = ØA εA + εD

Ø DØET

1 − ØET 1 − ØD( )

ØET=

I Aε

DØD− I D

εA ØA

εDØD I A

+ I DØA( )=

IA ØD − I DεA DØ A

ØD IA+ ID ØA( )

15

Where ØD , ØA ,Ø ET are the quantum yields for unsensitized donor and acceptor

fluoresence and energy transfer, respectively. εD , ε A , εA D are the extinction coefficients

of the donor and the acceptor and their ratio at the excitation wavelength. ID , IA are the

corrected measured intesities of the donor and the acceptor fluorescence from the single

molecule trajectories. Assuming the Förster mechanism for the energy transfer[34], the

yield ETØ for a distance R between chromophores is:

16

0ET R

R1Ø

−

+=

where R0 is the distance between chromophores that gives a quantum yield for energy

transfer of 50%. The factor R0 depends on the spectral overlap of donor emission and

acceptor absorption and on the angular part of the transition dipole-dipole interaction.

Distributions of energy transfer efficiency

For a given time point in a trajectory we average 2n+1 points (n before and n

after, in addition to the point) to obtain an observed count rate for the donor and the

acceptor. The number of points averaged determines the time resolution of the

distribution. We use n=1 for 3ms, n=12 for 25ms, and n=50 for 100ms time resolution

when measuring the distributions of energy transfer efficiency and deducing the donor-

acceptor distance distribution.

In Figure 8 the results of 60-70 trajectories are averaged to compute an overall

distribution of energy transfer efficiency for each concentration of denaturant. The

distribution is narrow and peaked at ~97% for 0M urea. The distribution broadens

considerably and the peak moves to ~90% for the midpoint of the urea denaturation curve

at 3M Urea. At 7.4M urea the peak of the distribution is at ~80% and further broadens.

16

Position distributions

Figure 9 shows the distributions PF(R) and PU(R) of the average donor/acceptor

distances for the folded (F) and unfolded (U) states calculated assuming an R-6 distance

dependence of the energy transfer rate and an angular factor k2 = 2/3. The function PF(R)

is obtained by subtracting 10% of PU(R) from the 0M urea distribution. As the urea

concentration increases, the average distance between chromophores and the variance of

the distribution calculated with these assumptions increases. Similarly the overall width

of the distance distribution for individual molecules increases with denaturant

concentration (data not shown). The peak of the distribution for PU(R) occurs at 25-30 Å,

which is somewhat longer than the value expected from the design. Models of GCN4-Pf

based on the crystal structure of GCN4-P1[15] would predict a distance of approximately

15-20 Å. This discrepancy is possibly due to unwinding the ends of the helices.

Furthermore there are systematic errors in the calculation of the energy transfer efficiency

due to the distribution of θ angles.

Discussion

Relationship of single molecule measurements to bulk measurements

The energy transfer quantum yield distributions in Figure 8 suggest that the

protein is most ordered at 0M urea, and less ordered at 7.4M urea, where the distribution

is broader, with 3M urea being intermediate between the two. This conclusion is

consistent with our understanding of the bulk folding experiments and directly leads to

properties of the folded and unfolded states of the protein. At pH 6.1 the protein is not

under completely native conditions. Therefore, even at 0M urea some fully unfolded

structures will be populated, assuming that the peptide folds in an effectively two-state

process.

17

We have considered the effect of the angular distribution on the energy transfer

distribution. From the fluorescence anisotropy measurements we showed that the

rotational motions of the donor and acceptor are nanosecond processes. In the single

molecule experiment, each photon we detect arises from a particular separation, R, and

orientation, Ω, of the donor and acceptor. The energy transfer efficiency was measured

by averaging ca. 6 photons per ms. So that at the shortest time resolution of 3 ms we

collect on average 18 photons. If we assume that R and Ω are statistically independent

and accept that the rotational motions occur over the whole range of Ω and are much

faster than the measurement, even 18 measurements approximately averages the angular

factor of the Förster theory to 2/3. This averaging should be much more complete for the

100 ms time resolution distributions. We will also assume that the signal is effectively

averaged over the colatitudes θA and θD that were discussed earlier as a possible source of

signal fluctuations. In future experiments, we expect to sharpen our knowledge of these

effects by analyzing the signals in polarized light. With the assumption of rotational

averaging we can convert the energy transfer efficiency distributions into the distance

distributions in Figure 9. The overall similarity of the distributions at different time

resolutions shows that the distance distribution is not completely averaged on the

measurement timescale.

The low ØET tail present in all the probability distributions increases in amplitude

as the denaturant concentration is increased. This tail is also present in the distributions

calculated from unaveraged trajectories from single molecules. Since individual

molecules can exhibit the entire range of configurations, the distributions (see Figure 8)

averaged over many trajectories must have a dynamical component.

18

Possible models that explain observed distributions

The bulk equilibrium may be described in terms of two states, one with a large

and the other with a small degree of helicity. However the single molecule experiments

are not expected to exhibit fluctuations between only two values of the energy transfer

yield. A range of energy transfer efficiencies would be a natural consequence of different

configurations being created each time the protein unfolds on the silanized glass surface.

In addition, there will be fluctuations of the structure within each thermodynamic state

that will cause energy transfer changes.

Our results clearly show that the distribution of energy transfer efficiencies and of

donor acceptor distances is quite different in 0M and 7.4M urea. We will refer to these

distributions as the folded and unfolded distributions. They may be close to the

equilibrium distributions for the two states although we would expect that there are also

rapid dynamic processes that are averaged on the ms time scale. Furthermore we obtain

the distribution function of R and we do not know the relationship between R and the

other coordinates of the protein. Nevertheless a picture of folding in this limited space

emerges, in which we have two states with quite different “folded” and “unfolded”

potentials of mean force, VF(R) and VU(R), corresponding to the two probability

distributions, PF(R) and PU(R), that we measured (see Figure 10). These potentials

incorporate the effects of the surface to which the protein is bound.

Our data also suggest that the rates of folding and unfolding depend on the donor

acceptor distance. If we adjust the urea concentration to the midpoint of the denaturation

curve the distribution observed should depend upon the relationship between the folding

time and the observation time required for the measurement. When the observation time

is much less than the folding-fluctuation time, the distribution is expected to be a linear

19

sum of the folded and unfolded distributions. The weights of folded and unfolded

components of the distribution should be equal if the molecule’s environment is exactly

at the midpoint of the titration curve.

Examination of the distribution observed in 7.4M urea with averaging times of 3,

25, and 100 ms (bottom panel) are uniformly broad. Thus, there is a broad distribution of

surface-associated unfolded states with different end-to-end distances that interconvert

slowly on the 100 ms time scale. Similarly, as the averaging time varies for the 0M urea

sample, the distribution shows a relatively small change. To examine the rates of

interconversion of folded and unfolded ensembles we examined linear combinations of

the 0 and 7.4M distributions, to determine whether they accurately predict the distribution

observed at 3M urea. Clearly as the averaging time changes from 3 to 100 ms, the fit to

the data is increasingly poor. Thus, some interconversion of the unfolded and folded

ensembles appears to occur on the 3-25 ms time scale. More specifically, the solid line in

the center panel of figure 8 is the linear combination of 50% of the 0M distribution and

50% of the 7.4M distribution. For the 100 ms time resolution data, the 3M distribution

does not have enough probability density at high ØET to fit directly to any linear

combination of the 0M and 7.4M distribution functions. Reducing the averaging time to

25ms improves the agreement of the linear combination of 0M and 7.4M to the 3M data

particularly in the tail of the distribution which now seems consistent with the two-state

model distribution. However the distributions still do not match well in the 90% region.

This implies that the fluctuations resulting in the differences in the tail of the distribution

observed at 100ms averaging time are, on average, slower than 25ms. Reducing the

averaging time to 3 ms results in a much better agreement between the 3M urea

20

distribution function and the linear combination of the 0M and 7.4M distribution

functions. We could obtain a good least squares fit of the 3M data to a linear combination

of the 0M and 7.4M data with weights 0.46 and 0.54 respectively. We conclude from the

results that fluctuations which are changing the energy transfer efficiency from ~97% to

~90% when changing the urea concentration from 0 to 3M occur on a time scale which is

between 3ms and 25ms. The low ØET portion of the distribution at 25 ms resolution still

matches well the higher resolution data, implying that fluctuations involving those values

of ØET are still slower than 25 ms. This result shows that there is a distribution of time

scales for fluctuations in the energy transfer efficiency and a correlation between kinetics

and structure. The time scales are structurally correlated such that faster time scales

correspond to fluctuations that modulate ØET in the 90% region while the slower time

fluctuations modulate ØET in the 70% and below region. The fluctuations that result in

low ØET values (<80%) are occurring on a time scale that is slower than 25ms.

Fluctuations in the ØET ~90% region are occurring faster than 25ms.

The separation of time scales established in this work suggests models for the

GCN4-Pf/silanized glass folding kinetics. A broad distribution of folding/unfolding rates

connecting two potentials VF(R) and VU(R) is indicated. Larger R values correspond to

slower rates. The result is that the waiting times between folding and unfolding events is

not exponentially distributed. This latter point was experimentally verified and will be

presented in a future publication along with the time correlations between donor and

acceptor emission [35].

This work illustrates the potential of single molecule experiments to measure the

distribution of molecular properties in a fluctuating and/or heterogeneous system. Not

21

only could the degree of heterogeneity be determined, but also the nature and time scales

of the associated fluctuations.

References1 X. S. Xie and J. K. Trautman, Annu. Rev. Phys. Chem. 49 (1998) 441-480.2 T. Funatsu, Y. Harada, M. Tokunaga, K. Saito, and T. Yanagida, Nature 374 (1995)

555.3 E. Geva and J. L. Skinner, Chem. Phys. Lett. 288 (1998) 225.4 T. Ha, A. Y. Ting, J. Liang, W. B. Caldwell, A. A. Deniz, D. S. Chemla, P. G.

Schultz, and S. Weiss, Proc. Natl. Acad. Sci. U. S. A. 96 (1999) 893-898.5 A. Ishijima, H. Kojima, T. Funatsu, M. Tokunaga, H. Higuchi, H. Tanaka, and T.

Yanagida, Cell 92 (1998) 161.6 H. P. Lu, L. Xun, and X. S. Xie, Science 282 (1998) 1877-1882.7 D. M. Warshaw, E. Hayes, D. Gaffney, A.-M. Lauzon, J. Wu, G. Kennedy, K.

Trybus, S. Lowey, and C. Berger, Proc. Natl. Acad. Sci. U. S. A. 95 (1998) 8034-8039.

8 L. M. Ying and X. S. Xie, J. Phys. Chem. B 102 (1998) 10399.9 Y. Jia, S. Sytnik, L. Li, S. Vladimirov, B. S. Cooperman, and R. M. Hochstrasser,

Proc. Nat. Acad. Sci. USA (1997)10 R. Zwanzig, Proc. Natl. Acad. Sci. U. S. A. 94 (1997) 148-150.11 W. A. Eaton, V. Munoz, Thompson, P. A. Henry, and J. E. R. Hofrichter, Acc. Chem.

Res. 31 (1998) 745.12 K. J. Lumb, C. M. Carr, and P. S. Kim, Biochemistry 33 (1994) 7361.13 H. Wendt, C. Berger, A. Baici, R. M. Thomas, and H. R. Bosshard, Biochemistry 34

(1995) 4097-107.14 N. D. Socci, J. N. Onuchic, and P. G. Wolynes, Proteins: Struct. Funct. Genet. 32

(1998) 136.15 E. K. O'Shea, J. D. Klemm, P. D. Kim, and T. Alber, Science 254 (1991) 539.16 T. E. Ellenberger, C. J. BRANDL, K. STRUHL, and S. C. HARRISON, Cell 71

(1992) 1223-1237.17 T. R. Sosnick, S. Jackson, R. R. Wilk, S. W. Englander, and W. F. DeGrado, Proteins

24 (1996) 427-32.18 H. Wendt, L. Leder, H. Harma, I. Jelesarov, A. Baici, and H. R. Bosshard,

Biochemistry 36 (1997) 204-13.19 H. Wendt, A. Baici, and H. R. Bosshard, J. Amer. Chem. Soc. 116 (1994) 6973-74.20 J. A. Zitzewitz, O. Bilsel, J. Luo, B. E. Jones, and C. R. Matthews, Biochemistry 34

(1995) 12812-9.21 I. Jelesarov, E. Durr, R. M. Thomas, and H. R. Bosshard, Biochemistry 37 (1998)

7539-50.22 E. Durr, I. Jelesarov, and H. R. Bosshard, Biochemistry 38 (1999) 870-80.23 S. Ozeki, T. Kato, M. E. Holtzer, and A. Holtzer, Biopolymers 31 (1991) 957-66.24 E. K. O'Shea, R. Rutkowski, and P. S. Kim, Science 243 (1989) 538-42.25 J. P. L. J. D. D. W. F. Schneider, J. Amer. Chem. Soc. 119 (1997) 5742.

22

26 G. B. Fields, Z. Tian, and G. Barany, in Synthetic Peptides A User's Guide, edited byG. A. Grant (W. H. Freeman and Company, 1992), pp. 126.

27 G. H. H. Olah, J. Chem. Phys. 89 (1988) 2531.28 H. P. Lu and X. S. Xie, Nature 385 (1997) 143.29 J. K. Trautman, J. J. Macklin, L. E. Brus, and E. Betzig, Nature 369 (1994) 40.30 T. Ha, T. Enderle, D. S. Chemla, P. R. Selvin, and S. Weiss, Phys. Rev. Lett. 77

(1996) 3979.31 T. Ha, T. Enderle, D. S. Chemla, P. R. Selvin, and S. Weiss, Chem. Phys. Lett. 271

(1997) 1.32 T. Basche, S. Kummer, and C. Brauchle, Nature 373 (1995) 132.33 M. A. Bopp, Y. Jia, G. Haran, E. A. Morlino, and R. M. Hochstrasser, Appl. Phys.

Lett. 73 (1998) 7-9.34 L. Streyer, Ann. Rev. Biochem. 47 (1978) 819-846.35 D. S. Talaga, W. L. Lau, Y. Jia, H. S. M. Lu, W. DeGrado, and R. M. Hochstrasser,

"Time Correlations Associated with Single Peptide Folding/Unfolding Dynamics"(To be published).

23

Figure Captions

Figure 1. Schematic representation of the folding of GCN4-Pf. The right panel shows the

crystal structure of folded GCN4-P1 with a hypothetical unfolded structure at the left.

The peptide adheres to the positively charged surface by electrostatic interaction with

the negatively charged glutamic acids at the C terminus of the peptide.

Conformational fluctuations cause changes in the donor-acceptor distance resulting in

an anticorrelated modulation in the donor and acceptor fluorescence intensities.

Figure 2. CD spectra of GCN4-Pf in solution (2.0 mM), and adsorbed on modified quartz

(surface density approximately 8 nm2/molecule). The buffer is 10 mM MES pH 6.1.

Figure 3. Urea unfolding curve for surface-adsorbed GCN4-Pf. The mean residue

ellipticity (MRE) at 222 nm was measured as a function of urea concentration, as

described in Experimental. The surface density of the peptide on the quartz surface is

approximately 1 molecule per 8 nm2.

Figure 4. Images of the donor (left) and acceptor (right) fluorescence of GCN4-Pf in the

absence of denaturant at pH 6.1. Notice that the acceptor channel is much brighter

than the donor channel.

Figure 5. Images of the donor (left) and acceptor (right) fluorescence taken at a urea

concentration of 7.4M at pH 6.1. The addition of denaturant has increased the donor

channel intensity markedly with respect to the acceptor channel.

24

Figure 6. Typical single molecule trajectories. The signals in the donor and acceptor

channels are labeled R6G and TxR, respectively. The trajectories (a), (b) and (c) are

discussed in the text.

Figure 7. High time resolution detection of donor (solid line) and acceptor (dashed line)

fluorescence signals from a single GCN4-Pf molecule at pH 6.1

Figure 8. Probability of occurrence P( ETØ ) of energy transfer efficiency ETØ versus

ETØ . The three rows correspond to urea concentrations 0, 3 and 7.4M. The three

columns correspond to averaging times of 3, 25 and 100 ms. The solid line in the 3M

urea distributions corresponds to an average of the 0 and 7.4M distributions for that

time resolution.

Figure 9. Probability distributions PF(R) and PU(R) of the average donor acceptor

separation, R, for the folded (F) and unfolded states (U).

Figure 10. Potentials of mean force for folded, VF(R), and unfolded, VU(R), states of

GCN4-Pf on silanized glass. The potentials were calculated from V(R) = -kBΤ ln

P(R), where P(R) is the relevant experimentally determined probability distribution of

donor acceptor distances.

25

Acknowledgements

This work was supported by GM54616 (to WFD), GM12592 (to RMH) and

GM48130 (to WFD and RMH) with instrumentation developed under RR01348.

D. Talaga was supported by NIH NRSA F32-GM18589.

26

Fig. 1

27

-50

0

50

100

150

200 220 240 260 280 300

[MR

E]

222

(10

3de

g cm

2dm

ol-1

)

28

-30

-25

-20

-15

-10

-5

0

-1 0 1 2 3 4 5 6

[MR

E] 22

2 (

103de

g cm

2dm

ol-1)

29

2 µ

30

2 µ

31

32

33

34

35

36